How long should a deep work session be?

The short answer

There is no universal optimum. The best-supported claim in the whole debate is much less exciting than any of the numbers: uninterrupted time beats fragmented time, and the switch back into work is more expensive than people think. A 50-minute block you actually finish is worth more than a 90-minute block you abandon twice.

If you want a starting point rather than a philosophy: begin at 45–60 minutes, and let two weeks of honest data move you up or down. The rest of this page explains why that's the starting point and not the answer.

Where each number comes from

25 minutes — the Pomodoro Technique

Francesco Cirillo's method, developed in the late 1980s, pairs a 25-minute work interval with a short break. Its 25 is not a finding about attention span; it's an interval Cirillo chose because it was short enough to commit to and long enough to be useful. That's a genuine strength — the Pomodoro's real mechanism is lowering the cost of starting, which is the hardest part of a work session for most people.

Where it strains is complex work. Twenty-five minutes is often just about the point where a hard problem stops feeling hostile, and a timer that ends there can cut you off at the moment the session was becoming worth it.

52 minutes — the 52/17 claim

This one circulates as though it were a laboratory result. It isn't. It comes from a productivity-tracking company's analysis of its own users' logged computer activity, published on its blog — an interesting observation about which of its users were most productive by its own measure, not a controlled experiment, not peer-reviewed, and not sampled from anything resembling the general population. Treat 52/17 as a plausible anecdote, not a law. Its useful contribution is directional: the most productive people in that dataset took real breaks, away from the screen, rather than none.

90 minutes — ultradian rhythms

The 90-minute figure traces back to Nathaniel Kleitman's basic rest–activity cycle: during sleep, the brain moves through roughly 90-minute cycles. That part is solid, decades-old sleep science. The extension — that the same 90-minute cycle continues through the waking day and governs when you can concentrate — is a hypothesis, and the waking evidence is considerably weaker and less consistent than the popular versions of it suggest.

So: 90 minutes is a reasonable ceiling for a single block, and many people do report a natural fade somewhere in that neighbourhood. But if someone tells you your brain runs on a strict 90-minute clock, they are stating a hypothesis as a fact.

What the research does support

• Switching costs are real and lingering. Sophie Leroy's work on attention residue (2009) found that when people move between tasks, part of their attention stays with the previous one — especially when it was left unfinished. This is the strongest argument against chopping deep work into small pieces.
• Sustained attention degrades over time. The vigilance decrement — performance falling over a continuous period of watchful attention — is one of the better-replicated findings in attention research. It argues for an upper bound on block length, though the classic studies use monotonous monitoring tasks, not engaging creative work, so don't map their timings onto your writing session.
• Breaks help, and their quality matters. Brief detachment from the task tends to restore performance. A break spent scrolling a feed is not the same intervention as a break spent walking, staring out of a window, or making coffee.
• Task type dominates. Shallow, well-defined work tolerates short blocks. Work with a long warm-up — debugging, drafting, mathematics, anything where you hold a lot of state in your head — pays a much higher restart tax and rewards longer blocks.

Notice what's missing from that list: any evidence for a specific magic number. The literature supports the shape — long enough to get in, short enough to stay sharp, with real breaks — and leaves the exact minutes to you.

Match the block to the work

A more useful question than "how long is a deep work session" is "how long is the warm-up for this task?" Rough guide, from our own sessions and worth exactly what a rough guide is worth:

• Admin, email, code review, tidying: 25–30 minutes. Short warm-up, easy to resume, and the timer's pressure is a feature.
• Reading, studying, learning something new: 45–60 minutes. Long enough to get past the first uncomfortable stretch where nothing sticks.
• Writing, coding, design, hard problem-solving: 60–90 minutes. High restart tax; protect these blocks harder than the others.
• Anything you're avoiding: 15 minutes. Optimise for starting, not for depth. You can always keep going, and usually will.

Sound choice follows the same logic — a session you intend to hold for 90 minutes wants audio without novelty in it, which is why lyrics and shuffling playlists tend to fail long blocks. We go into that in our guide to music for deep work, and into the frequency side in beta waves and 18 Hz.

Find your number in two weeks

1. Pick one recurring task type — the same kind of work each time, so you're comparing like with like.
2. Week one: run 45-minute blocks. One variable only. Record three things at the end of each block: did you finish it without escaping, how long before the first pull toward your phone, and whether the work was any good.
3. Week two: adjust once. If you were consistently still sharp at the bell, go to 75 or 90. If you broke early most days, drop to 30 and rebuild. Don't change length daily — you'll only be measuring noise.
4. Keep the break honest. Same break every time, away from the screen. A break that is just a different screen is not a break, and it will quietly contaminate the whole experiment.
5. Judge on completed blocks, not intended ones. The number you want is the longest block you reliably finish — not the longest one you can imagine finishing.

Why we count completed sessions only

That last point is the one we built into the product. NeuroWave's Focus Receipts summarize sessions you actually completed — not sessions you started, not sessions you meant to have. A tool that congratulates you for a 90-minute block you bailed on at minute twelve is teaching you the wrong thing, and it makes the two-week test above impossible to run honestly.

It's also why the app doesn't hand you a receipt until there's enough real history to say anything meaningful. Three sessions is a mood; a fortnight of finished blocks is a number you can plan around. If you want the fuller picture of how we compare to the alternatives, our compare pages lay it out plainly, and the rest of the Learn library covers the sound side in more depth.